Research On Robotic Arm Grasping Technology For Multi-Types Moving Object

With the advent of Industry 4.0,new technologies such as robotic arms and computer vision are highly integrated,and today,conveyor belts are widely used in factory distribution,warehousing,manufacturing and production for automation and faster delivery.How to combine technologies such as visual tracking and object grasping so that the robotic arm can successfully grasp the objects on conveyor is a challenge.In order to improve the grasping efficiency and accuracy,the grasping of multi-types moving objects is studied in this paper,and the specific work is as follows:(1)For the object detection problem of the robotic arm,a light-weight network grasping detection method is proposed.The YOLOv4 network is used as the main frame to improve the information acquisition of the moving target.In order to meet the real-time requirements of network running on the robotic arm,this paper improves the YOLOv4 target detection algorithm,replaces the backbone network with the Moile Net V3 network,and reduces the amount of network computation.The average detection accuracy on the Cornell grasping dataset can reach 96.68%,and the detection speed is 2.28/ms,which can meet the requirements of the robotic arm for grasping moving objects.For the grasping detection of the gripper,a multi-scale oriented anchor box mechanism adapted to the grasping characteristics of the robotic arm is introduced for grid prediction,which improves the accuracy of grasping multi-types of objects.(2)For the moving target on the conveyor belt,the calculation method of the speed of the conveyor belt and the pose estimation of the moving target are studied.A moving object prediction network combining Moile Net V3-YOLOv4 grasping detection algorithm with LSTM(Long Short Term Memory)network is proposed to predict the position information of moving objects.When the robotic arm grasps,the trajectory planning in the joint space and the "gate-shaped" path planning method of the robotic arm are used to improve the grasping efficiency.Through experiments,it is verified that the method used in this paper can accurately predict the target position and complete the grasping task.Finally,on the grasping platform of the robotic arm,the tracking and grasping experiment of the moving target on the conveyor belt is carried out.The results show that when the speed of the conveyor belt is 100mm/s,the grasping success rate of the robotic arm is 97%,the average grasping time is 2.806 s.For objects with different speeds,it can also effectively complete the task of grabbing the target and meet the expected results.